As it turns out, because the kernel matrix for kernel ridge regression has all positive values and is symmetric, it's possible to use a specialized matrix inverse algorithm called Cholesky ...